Beruflich Dokumente
Kultur Dokumente
Received: August 11, 2011 / Accepted: November 08, 2011 / Published: August 31, 2012.
Abstract: This paper presents an ANN (artificial neural networks)-based technique for improving the performance of distance relays
against open-circuit faults in transmission networks. The technique utilizes the small capacitive current measured in the open-phase
plus the currents in the two healthy phases in calculating the open-circuit fault distance. The results obtained show that a distance relay
with the proposed scheme will not only be able to detect the open-conductor condition in HVTL (high voltage transmission line) but
also to locate the place of this fault regardless the value of the pre-fault current loading. There is no need for especial communication
schemes since the existing media could work properly for the needs of the proposed technique.
Key words: Distance relay, open-conductor, ANN, transmission networks, fault location.
it. The protection scheme which may detect this type of 2. Typical Problems
fault (according to the pre-fault current value) is the
For an open-circuit fault occurred in a single phase
earth fault relay. However, even if the line is equipped
of line C-D as shown in Fig. 1, the distance relays R1,
with an earth fault protection and the pre-fault current
R2 and R3 can easily detect such unsymmetrical fault
value was large enough to initiate the operation of the
since one of the three phases is almost zero. However,
earth fault relay; there will be a large time delay since
unless the settings for the three relays (as well as the
earth fault relays is actually a backup protection for
settings of the associated earth fault relays) are
HVTLs (high voltage transmission lines).
accurately estimated, a discrimination problem
There is a lack of researches considering cases of
between the three relays is expected and a false trip
open (downed) conductor, since most of protective
from relays R1 and R2 can not be avoided. Usually, the
relays do not respond to these types of faults. Protection
open-circuit case starts with a simple problem and ends
engineers believe that the protective devices are not
with a wide disconnected area.
designed for open-conductor fault case. Even the field
A recorded open-circuit fault case is illustrated in
experts find that no response from protective relays for
Fig. 2. An open-circuit fault occurred on one of the two
case of open-conductor is accepted. There are not
parallel circuits. The fault, as expected, was not
significant research efforts in the field of using distance
detected by the distance relays at both sides of the line.
relay to detect or locate the open-conductor fault Instead, it was detected by the earth fault relays: EFR-1
especially in the high voltage interconnected systems. and EFR-2.
The only effort exerted in field of studying The unbalance level at the earth fault relay EFR-2 (at
open-conductor problems is in distribution networks 220 kV side) is much higher than the unbalance level
[1, 3-5]. A brief review of some open-conductor fault at EFR-1 (at 66 kV side). Consequently, the earth fault
detection techniques can be found in Refs. [6-8]. Most relay EFR-2 tripped faster than the earth fault relay
of the proposed techniques are based on a device
located in the substation, such as a relay which operates
R1 R1' R2 R2' R3 R3'
monitoring non-60 Hz parameters of the feeder current
[9, 10] as changes in second or third harmonics for
Fig. 1 Open circuit fault.
example.
In all the previous studies, the research work is
concentrating on detecting the open-conductor fault
case. However, there is no work that is done in the area
of estimating the distance of the open-circuit fault. The
objective of this paper is to modify the typical distance
relay algorithm to be able to estimate not only the
distance to short-circuit-faults but also to detect and
estimate the distance to open-circuit faults. In this work,
the line is assumed to be open from both sides. In the
next sections, the effect of open-circuit faults compared
with short-circuit faults is demonstrated with typical
recorded case studies. The mathematical formulation
of open-circuit problem and the proposed ANN-based
technique are then presented. Fig. 2 Reverse grading problem.
Novel Algorithm for Estimating the Distance of Open-Conductor Faults in HV Transmission Lines 1303
4. Problem Formulation
SLM 66 kV
Fig. 3 Wide disconnected problem resulted from Considering the power system shown in Fig. 4, a
open-conductor. single-phase open-conductor fault (in phase A) is
1304 Novel Algorithm for Estimating the Distance of Open-Conductor Faults in HV Transmission Lines
open-circuit (Z1 and Z0). These values are the input values fed to the ANN
The last factor can not be obtained by any distance where the output is the required positive sequence
relay since it is considered as unforeseen data. The impedance of the line till the place of open-conductor.
proposed distance relay in this paper is using Eq. (8) to
6. Simulation Results
estimate the location of the open-conductor using
ANN. The proposed scheme identifies the case of
open-conductor based on the level of current in each of
5. ANN Configuration the three phases. Under single phase open-conductor
ANN is a suitable technique to be used in this study condition, the current in one of the three phases will be
since ANNs are prepared to work with incomplete and below a current threshold (Idetc). The scheme also
unforeseen data. checks if the phase breaker is open or not (to avoid
The researchers were motivated to use ANN for conflict with the case of segregated phase breakers).
transmission line relaying for several reasons [13-15]. The case of open-conductor is confirmed if the
First, the ability of ANNs to learn complex nonlinear difference between the other two phase currents is less
input/output relations and their inherent immunity to than an error.
changing operating conditions played a role. At the Different open-circuit faults were applied at
same time, ANNs have a high degree of robustness and different locations along the line X-Y in the power
ability to learn. system model shown in Fig. 4. Part of the power
In this paper, different configurations (different system specifications is presented in Table 1.
number of hidden layers, different number of neurons in Fig. 5 shows typical results for an open-circuit case
hidden layer, different transfer function for hidden and at distance of 10% of the protected line measured from
output layer(s) and different training methods) are tested the relay RX. The calculated RMS values for current and
in order to choose the best configuration for the voltage signals at both sides of the line are shown in
proposed ANN scheme. The configuration achieved Figs. 5a-5d where the impedance seen by distance
minimum error goal includes two hidden layers with relays RX and RY are shown in Figs. 6a and 6b.
five neurons. The hidden layer neurons transfer function It is clear that the relay can not detect the open fault
is Tansig where the output layer neuron transfer function case since the impedance seen by the relay is much
is Puerlin. The proposed ANN based fault classification larger than the load impedance.
scheme has been designed using Trainscg function. Table 2 shows a summary of test results of the
MatLab is used in building and training the ANN. proposed scheme. The maximum error recorded is less
The values of the sequence currents versus the than 11%. This can be improved by training the ANN
positive sequence impedance Z1 are analyzed with more fault cases. The error is accepted taking into
statistically to get the best formation of the training account that the ordinary distance protection is not able
inputs. This yields the following three inputs: to even detect such kind of faults.
I 7. Conclusions
Input 1 1PF
I1 I 2
2 A novel ANN-based technique is proposed to
I
Input 2 2 (9) improve the performance of the ordinary distance relay
I0 against open-circuit faults. The maximum error in the
2
I calculated impedance is less than 11%. The results
Input 3 1
I2 obtained show that a distance relay with the proposed
1306 Novel Algorithm for Estimating the Distance of Open-Conductor Faults in HV Transmission Lines
Z (at side X)
Equipment Data 40000
30000
Feeding source at side-X 2.5 GVA
20000
Feeding source at side-Y 3.05 GVA
10000
Line length 100 km
0
Capacitance 3.8 × 10-6 F/km
0.00 0.05 0.10 0.15 0.20
Positive sequence reactance 0.296 Ω/km Time
Positive sequence resistance 0.04 Ω/km
(a) Calculated impedance at side X
Zero sequence reactance 0.25 Ω/km
4000
Zero sequence resistance 0.95 Ω/km
Z (at side Y)
3000
2000
2000
1000
IA (side X)
1500
0
1000
0 0.05 Time 0.1 0.15 0.2
500
0
(b) Calculated impedance at side Y
0 0.05 0.1 0.15 0.2
Fig. 6 Calculated impedance at both ends under
Time open-circuit fault condition.
could work properly for the needs of the proposed Downed Conductors on Utility Distribution Systems,
Institute of Electrical and Electronics Engineers, New
technique.
York, Feb. 1990.
[8] Downed Power Lines: Why They Can not Always Be
Acknowledgments
Detected, IEEE Power Engineering Society Public Affairs
The financial support from the KFAS (Kuwait Document, Feb. 1989 (Green book).
[9] C.L. Benner, B.D. Russel, Practical high impedance fault
Foundation for Advancement of Science) is highly
detection on distribution feeders, IEEE Transactions on
appreciated. Industry Applications 33 (3) (1997) 635-640.
[10] H. Calhoun, M.T. Bishop, C.H. Eichler, R.E. Lee,
References Development and testing of an electro-mechanical relay to
[1] E.C. Senger, Broken conductor protection system using detect fallen distribution conductors, IEEE Transactions
carrier communication, IEEE Transactions on Power on Power Apparatus and Systems PAS-101 (6) (1982)
Delivery 15 (2) (2000) 525-530. 1643-1648.
[2] B.M. Weedy, B.J. Cory, Electrical Power Systems, 4th ed., [11] ALSTOM, Technical Guide Distance Protection Relays,
John Wiley & Sons, New York, 1998. TG 1.1671-B, Vol. 1, pp. 441-444.
[3] IEEE Guide for Determining Fault Location on AC [12] Power System Protection, The Electricity Council, UK,
Transmission and Distribution Lines, IEEE Standard Vol. I, 1989.
C37.114, Dec. 2004. [13] R.N. Mahanty, P.B.D. Gupta, Comparison of fault
[4] C. Fan, K.K. Li, W.L. Chan, W. Yu, Z. Zhang, classification methods based on wavelet analysis and
Application of wavelet fuzzy neural network in locating ANN, Electrical Power Components and Systems 34 (1)
single line to ground fault (SLG) in distribution lines, (2006) 47-60.
Electrical Power and Energy Systems 29 (2007) 497-503. [14] D.V. Coury, D.C. Jorge, Artificial neural network
[5] D.I. Jeerings, J.R. Linders, A practical protective relay for approach to distance protection of transmission lines,
down-conductor faults, IEEE Trans. on Power Delivery 6 IEEE Trans. Power Delivery 13 (1) (1998) 102-108.
(2) (1991) 565-574. [15] R. Venkatesan, B. Balamurugan, A real time hardware
[6] High Impedance Fault Detection Technology, Report of fault detector using an artificial neural network for
PSRC working group D15, Mar. 1996. distance protection, IEEE Trans. on Power Delivery 16 (1)
[7] B.D. Russell, IEEE Tutorial Course Text, Detection of (2001) 75-82.